FitzGerald J., Dennis A., Durcikova A. Business Data Communications and Networking
Подождите немного. Документ загружается.
8.3 DEDICATED-CIRCUIT NETWORKS 275
Toronto
Vancouver
Dallas
Atlanta
New York
San Francisco
Los Angeles
FIGURE 8.3 Ring-based design
one circuit or computer becomes overloaded. A long delay in any one circuit or computer
can have significant impacts on the entire network.
In general, the failure of any one circuit or computer in a ring network means that
the network can continue to function. Messages are simply routed away from the failed
circuit or computer in the opposite direction around the ring. However, if the network is
operating close to its capacity, this will dramatically increase transmission times because
the traffic on the remaining part of the network may come close to doubling (because
all traffic originally routed in the direction of the failed link will now be routed in the
opposite direction through the longest way around the ring).
Star Architecture A star architecture connects all computers to one central com-
puter that routes messages to the appropriate computer (Figure 8.4). The star topology
is easy to manage because the central computer receives and routes all messages in
the network. It can also be faster than the ring network because any message needs to
travel through at most two circuits to reach its destination, whereas messages may have
to travel through far more circuits in the ring network. However, the star topology is
the most susceptible to traffic problems because the central computer must process all
messages on the network. The central computer must have sufficient capacity to handle
traffic peaks, or it may become overloaded and network performance will suffer.
276 CHAPTER 8 WIDE AREA NETWORKS
Toronto
Vancouver
Dallas
Atlanta
New York
San Francisco
Los Angeles
FIGURE 8.4 Star-based design
In general, the failure of any one circuit or computer affects only the one computer
on that circuit. However, if the central computer fails, the entire network fails because all
traffic must flow through it. It is critical that the central computer be extremely reliable.
Mesh Architecture In a full-mesh architecture, every computer is connected to
every other computer (Figure 8.5a). Full-mesh networks are seldom used because of the
extremely high cost. Partial-mesh architecture (usually called just mesh architecture),
in which many, but not all, computers are connected, is far more common (Figure 8.5b).
Most WANs use partial-mesh topologies.
The effects of the loss of computers or circuits in a mesh network depend entirely
on the circuits available in the network. If there are many possible routes through the
network, the loss of one or even several circuits or computers may have few effects
beyond the specific computers involved. However, if there are only a few circuits in the
network, the loss of even one circuit or computer may seriously impair the network.
In general, mesh networks combine the performance benefits of both ring networks
and star networks. Mesh networks usually provide relatively short routes through the
network (compared with ring networks) and provide many possible routes through the
8.3 DEDICATED-CIRCUIT NETWORKS 277
Toronto
Vancouver
Dallas
Atlanta
New York
San Francisco
Los Angeles
Toronto
Vancouver
Dallas
Atlanta
New York
San Francisco
Los Angeles
(a) Full mesh
(b) Partial mesh
FIGURE 8.5 Mesh design
network to prevent any one circuit or computer from becoming overloaded when there
is a lot of traffic (compared with star networks in which all traffic goes through one
computer).
The drawback is that mesh networks use decentralized routing so that each com-
puter in the network performs its own routing. This requires more processing by each
278 CHAPTER 8 WIDE AREA NETWORKS
computer in the network than in star or ring networks. Also, the transmission of network
status information (e.g., how busy each computer is) “wastes” network capacity.
There are two types of dedicated-circuit services in common use today: T carrier
services and synchronous optical network (SONET) services. Both T carrier and SONET
have their own data link protocols, which are beyond the focus of this chapter.
8.3.2 T Carrier Services
T carrier circuits are the most commonly used form of dedicated-circuit services in
North America today. As with all dedicated-circuit services, you lease a dedicated circuit
from one building in one city to another building in the same or different city. Costs are
a fixed amount per month, regardless of how much or how little traffic flows through the
circuit. There are several types of T carrier circuits as shown in Figure 8.6, but only T1
and T3 are in common use today.
A T1 circuit (also called a DS1 circuit) provides a data rate of 1.544 Mbps. T1
circuits can be used to transmit data but often are used to transmit both data and voice.
In this case, inverse TDM provides 24 64-Kbps circuits.
2
Digitized voice using PCM
requires a 64-Kbps circuit (see Chapter 3), so a T1 circuit enables 24 simultaneous voice
channels. Most common carriers make extensive use of PCM internally and transmit
most of their voice telephone calls in digital format using PCM, so you will see many
digital services offering combinations of the standard PCM 64-Kbps circuit.
A T3 circuit allows transmission at a rate of 44.736 Mbps although most articles
refer to this rate as 45 megabits per second. This is equal to the capacity of 28 T1
circuits. T3 circuits are becoming popular as the transmission medium for corporate
MANs and WANs because of their higher data rates. Although T2 and T4 circuits are
defined standards, they are not commercially available and therefore we don’t discuss
them here.
Fractional T1, sometimes called FT1, offers portions of a 1.544-Mbps T1 circuit
for a fraction of its full cost. Many (but not all) common carriers offer sets of 64 Kbps
DS-0 channels as FT1 circuits. The most common FT1 services provide 128 Kbps,
256 Kbps, 384 Kbps, 512 Kbps, and 768 Kbps.
T Carrier
Designation
DS Designation Speed
FT1
T1
T2
T3
T4
DS0
DS1
DS2
DS3
DS4
64 Kbps
1.544 Mbps
6.312 Mbps
44.376 Mbps
274.176 Mbps
FIGURE 8.6 T carrier
services
2
If you multiply 24 circuits by 64 Kbps per circuit, you will get 1.536 Mbps, not 1.544 Mbps. This is because
some of the 1.544-Mbps circuit capacity is used by the common carrier for control signals used to frame the
data (i.e., mark the start and stop of packets).
8.4 PACKET-SWITCHED NETWORKS 279
SONET
Designation SDH Designation Speed
OC-1
OC-3
OC-12
OC-24
OC-48
OC-192
STM-1
STM-4
STM-8
STM-16
STM-24
51.84 Mbps
155.52 Mbps
622.08 Mbps
1.244 Gbps
2.488 Gbps
9.953 Gbps
OC-768 STM-256 39.813 Gbps
OC-3072 STM-1024 159.25 Gbps
FIGURE 8.7 SONET
(synchronous optical network)
and SDH (synchronous digital
hierarchy) services. OC = optical
carrier (level)
8.3.3 Synchronous Optical Network
The synchronous optical network (SONET) is the American standard (ANSI) for
high-speed dedicated-circuit services. The ITU-T recently standardized an almost iden-
tical service that easily interconnects with SONET under the name synchronous digital
hierarchy (SDH).
SONET transmission speeds begin at the OC-1 level (optical carrier level 1) of
51.84 Mbps. Each succeeding rate in the SONET fiber hierarchy is defined as a multiple
of OC-1, with SONET data rates defined as high as 160 Gbps. Figure 8.7 presents
the commonly used SONET and SDH services. Each level above OC-1 is created by
an inverse multiplexer. Notice that the slowest SONET transmission rate (OC-1) of
51.84 Mbps is slightly faster than the T3 rate of 44.376 Mbps.
8.4 PACKET-SWITCHED NETWORKS
Packet-switched networks are quite different from the two types of networks dis-
cussed previously. For both circuit-switched and dedicated-circuit networks, a circuit was
8.1 CLEVELAND TRANSIT
MANAGEMENT
FOCUS
The Greater Cleveland Regional Transit Authority
(GCRTA) has about 2400 employees and provides
bus, trolley, and rail service to about 1.3 million
people in the Cleveland area. It has many office
locations throughout the region, and more than
half of its employees are on the move as they
work.
A recent blackout highlighted how vulnerable
GCRTA was to network outages. Communications
were knocked out, including systems supporting
the Transit police.
GCRTA redesigned its WAN to use a SONET
ring. SONET provides high-speed data services
and the ring topology ensure maximum reliability.
Even if one part of the ring is knocked out, whether
by power failures or someone accidentally cutting
a line, the network will continue to operate.
SOURCE: ‘‘Staying on Track,’’ Case Study, AT&T,
2008.
280 CHAPTER 8 WIDE AREA NETWORKS
established between the two communicating computers. This circuit provided a guaran-
teed data transmission capability that was available for use by only those two computers.
In contrast, packet-switched services enable multiple connections to exist simultaneously
between computers over the same physical circuit, just like LANs and BNs.
8.4.1 Basic Architecture
With packet-switched services, the user again buys a connection into the common car-
rier cloud (Figure 8.8). The user pays a fixed fee for the connection into the network
(depending on the type and capacity of the service) and is charged for the number of
packets transmitted.
The user’s connection into the network is a packet assembly/disassembly device
(PAD), which can be owned and operated by the customer or by the common carrier.
The PAD converts the sender’s data into the network layer and data link layer packets
used by the packet network and sends them through the packet-switched network. At
the other end, another PAD reassembles the packets back into the network layer and
data link layer protocols expected by the destination and delivers it to the appropriate
computer.
One of the key advantages of packet-switched services is that different locations can
have different connection speeds into the common carrier cloud. The PAD compensates
for differences in transmission speed between sender and receiver; for example, the
circuit at the sender might be 1.5 Mbps whereas the receiver only has a 64-Kbps circuit.
In contrast, a dial-up circuit or a dedicated circuit must have the same speed at both the
sender and receiver.
Packet-switched networks enable packets from separate messages with different
destinations to be interleaved for transmission, unlike switched circuits and dedicated
LAN
LAN
LAN
LAN
LAN
Common
carrier
network
PAD
PAD
PAD
PAD
PAD
FIGURE 8.8 Packet-switched
services. LAN = local area network;
PAD = packet assembly/disassembly
device
8.4 PACKET-SWITCHED NETWORKS 281
circuits. Packet switching is popular because most data communications consist of short
bursts of data with intervening spaces that usually last longer than the actual burst of
data. Packet switching takes advantage of this characteristic by interleaving bursts of
data from many users to maximize use of the shared communication network. Figure 8.9
shows a packet-switching connection between six different cities. The little boat-shaped
figures (shown on the communication circuits) represent individual packets from
separate messages.
Although the packets in one data stream may mix with several other data streams
during their journey, it is unlikely that packets from two different data streams will travel
together during the entire length of their transmission. The two communicating computers
do not need to know through which intermediate devices their data are routed because
the packet network takes care of it by either of two methods.
The first method, called datagram, is a connectionless service. It adds a destination
address and sequence number to each packet, in addition to information about the data
stream to which the packet belongs. In this case, a route is chosen for each packet as it
is accepted into the packet network. Each packet may follow a different route through
the network. At the destination address, the sequence number tells the network how to
reassemble the packets into a continuous message. The sequence number is necessary
because different routes may deliver packets at different speeds, so data packets often
arrive out of sequence. Few networks today use datagrams for data transfer.
The second and more common routing method is a connection-oriented approach
called a virtual circuit. In this case, the packet-switched network establishes what appears
to be one end-to-end circuit between the sender and receiver. All packets for that trans-
mission take the same route over the virtual circuit that has been set up for that particular
transmission. The two computers believe they have a dedicated point-to-point circuit, but
in fact, they do not.
Error
check
(16 bits)
Data
(up to
1024 bits)
Header
address,
control
What is in a Packet?
64 Kbps
1.5 Mbps
128 Kbps
64 Kbps
128 Kbps
1.5 Mbps
Packet switching network
Packet
Character
PAD (packet assembly/disassembly)
Legend:
City A
Node
City B
Node
FIGURE 8.9 Packet-switching concepts
282 CHAPTER 8 WIDE AREA NETWORKS
Virtual circuits are usually permanent virtual circuits (PVCs), which means that
they are defined for frequent and consistent use by the network. They do not change
unless the network manager changes the network. Some common carriers also permit the
use of switched virtual circuits (SVCs) although this is not common. Changing PVCs is
done using software, but common carriers usually charge each time a PVC is established
or removed.
Because most network managers build packet-switched networks using PVCs, most
packet-switched networks behave like dedicated-circuit networks. At first glance, the basic
architecture in Figure 8.8 looks very similar to the cloud mesh of switched-circuit
services, and in fact, they are very similar because data can move from any com-
puter attached to the cloud to any other on the cloud. However, because virtually all
data-intensive networks use PVCs, this means that the network is actually built using
virtual circuits that are the software equivalent of the hardware-based dedicated circuits.
As a result, packet-switched networks often have star, ring, and mesh topologies similar
to those of dedicated-circuit networks. However, because the PVCs are in software, they
are much easier to change than the hardware circuits of dedicated-circuit networks.
Most common carriers permit users to specify two different types of data rates that
are negotiated per connection and for each PVC as it is established. The committed
information rate (CIR) is the data rate the PVC must guarantee to transmit. If the
network accepts the connection, it guarantees to provide that level of service. Most
connections also specify a maximum allowable rate (MAR), which is the maximum
rate that the network will attempt to provide, over and above the CIR. The circuit will
attempt to transmit all packets up to the MAR, but all packets that exceed the CIR are
marked as discard eligible (DE). If the network becomes overloaded, DE packets are
discarded. So although users can transmit more data than the CIR, they do so at a risk
of lost packets and the need to retransmit them.
Packet-switched services are often provided by different common carriers than
the one from which organizations get their usual telephone and data services. There-
fore, organizations often lease a dedicated circuit (e.g., T1) from their offices to the
packet-switched network pointofpresence(POP).The POP is the location at which
the packet-switched network (or any common carrier network, for that matter) connects
into the local telephone exchange.
There are four types of packet-switched services: ATM, frame relay, IP/MPLS, and
Ethernet services. Several common carriers (e.g., Sprint) have announced that they intend
to stop offering all services except MPLS (Multi-protocol label switching), Ethernet,
and Internet services (see Chapter 9). Other carriers have hinted at the same decision.
Over the next few years these technologies may disappear.
8.4.2 Asynchronous Transfer Mode
Asynchronous transfer mode (ATM) works in much the same way as the Ethernet and
TCP/IP networks we have discussed in previous chapters. However, ATM uses different
layer-2 and layer-3 protocols. ATM has its own layer-3 protocol but typically uses SONET
at layer 2. ATM usually performs encapsulation, which means that packets that enter an
ATM network have their Ethernet frame preserved. This Ethernet frame is surrounded
(not replaced) by an ATM layer-3 packet and a SONET frame which are used to move
8.4 PACKET-SWITCHED NETWORKS 283
them through the ATM network. When the packet arrives at the destination and leaves
the ATM network, the ATM packet and SONET frame are stripped off and the original
Ethernet frame is delivered unchanged into the destination network.
Second, ATM provides no error control in the network; error control is the respon-
sibility of the source and destination. (ATM is considered an unreliable packet service.)
Because the user’s data link packet remains intact, it is simple for the devices at the edge
of the ATM network to check the error-control information in the packet to ensure that
no errors have occurred and to request transmission of damaged or lost packets.
Third, ATM provides extensive QoS information that enables the setting of very
precise priorities among different types of transmissions: high priority for voice and
video, lower priority for e-mail.
Finally, ATM is scalable; it is easy to multiplex basic ATM circuits into much
faster ATM circuits. Most common carriers offer ATM circuits that provide the same data
transmission rates as SONET: 51.84 Mbps, 466.56 Mbps, 622.08 Mbps, and so on up to
39 Gbps (OC-768). New versions called T1 ATM (1.544 Mbps) and T3 ATM (45 Mbps)
are also available.
8.4.3 Frame Relay
Frame relay, just recently standardized, transmits data slower than ATM so it has some-
times been called a poor man’s ATM. It is one of the most commonly used WAN services
in the world. Like ATM, frame relay performs encapsulation of packets, so packets are
delivered unchanged through the network. Like ATM, it is an unreliable packet service
because it does not perform error control. Frame relay checks for errors but simply dis-
cards packets with errors. It is up to the software at the source and destination to control
for lost messages.
Frame relay does not yet provide QoS capabilities, but this is under development.
Different common carriers offer frame relay networks with different transmission speeds.
Most offer a range of CIR speeds that include 56 Kbps, 128 Kbps, 256 Kbps, 384 Kbps,
1.5 Mbps, 2 Mbps, and 45 Mbps.
8.2 FRAME RELAY AT AIR CHINA
MANAGEMENT
FOCUS
Air China is the largest airline in China, both in
terms of traffic and assets. The airline has over
200 aircraft that serve 81 domestic and 42 inter-
national destinations, with about 6000 scheduled
flights per week. The airline industry has become
more competitive due to China’s rapid economic
growth, so the airline sought to improve is global
WAN.
Air China needed a strong global network to
link its headquarters in Beijing with its 40 offices in
Asia, Europe, the Middle East, Africa, and North
America. It partnered with AT&T to provide an
IP-based network running over frame relay in all
countries. For local equipment, they choose Cisco
routers and PADs to connect into the network.
SOURCE: ‘‘Managed Networking Solutions Offer
Competitive Advantage for China’s Largest Global
Airline,’’ Case Study, AT&T, 2008.
284 CHAPTER 8 WIDE AREA NETWORKS
8.4.4 Ethernet Services
Although we have seen rapid increases in capacities and sharp decreases in costs in LAN
and BN technologies, changes in WAN services offered by common carriers saw only
modest changes in the 1990s. That changed in 2000 with the introduction of several
Internet startups (e.g., Yipes) offering Ethernet services.
Most organizations today use Ethernet and IP in the LAN and BN environment, yet
the WAN packet network services (ATM and frame relay) discussed earlier use different
layer-2 protocols. Any LAN or BN traffic, therefore, must be translated or encapsulated
into a new protocol and destination addresses generated for the new protocol. This takes
time, slowing network throughput. It also adds complexity, meaning that companies must
add staff knowledgeable in the different WAN protocols, software, and hardware these
technologies require. This is one reason many common carriers are starting to call these
technologies “legacy technologies,” signaling their demise.
Each of the preceding packet services uses the traditional PSTN provided by the
common carriers such as AT&T and BellSouth. In contrast, Ethernet services bypass
the PSTN; companies offering Ethernet services have laid their own gigabit Ethernet
fiber-optic networks in large cities. When an organization signs up for service, the packet
network company installs new fiber-optic cables from their citywide backbone into the
organization’s office complex and connects it to an Ethernet switch. The organization
simply plugs its network into its Ethernet switch and begins using the service. All traffic
entering the packet network must be Ethernet, using IP.
A Day in the Life: Networking and Telecommunications Vice President
A vice president is a person in an executive-level posi-
tion whose focus is to set the strategic direction for the
organization. A vice president has very little to do with
the day-to-day operations; much like an Admiral in a
Navy fleet, he or she defines the direction, but the indi-
vidual captains running each ship actually make sure
that everything that needs to happen gets done.
The vice president works with the chief informa-
tion officer (CIO) and other executive leadership of the
organization to identify the key organizational goals
that have implications for the network. The vice pres-
ident works with his or her staff to revise the strategic
networking plan to ensure that the network is capable
of supporting the organization’s goals. The key ele-
ments of the strategic plan are the networking architec-
tures, key technologies, and vendors. Once the strategy
has been set, the vice president’s job is to instruct the
senior managers to execute the strategy and then let
them do their jobs.
In most cases, the changes to the networking strate-
gic plan are relatively minor, but sometimes there are
dramatic changes that require a major shift in strategic
direction. For example, in recent years, we’ve seen a
major change in the fundamental capabilities of net-
work tools and applications. Our architecture strategy
during the 1990s was driven by the fact that network
management tools were poor and maintenance costs per
server were high; the fundamental architecture strategy
was to minimize the number of servers. Today, network
management tools are much better, maintenance costs
per server are significantly lower, and network traffic
has changed both in volume and in the number and com-
plexity of services supported (e.g., Web, email, H.323,
IPv6); the strategy today is to provide a greater number
of servers, each of which is dedicated to supporting one
specific type of traffic.
With thanks to Brian Voss